Policy Design for an Ankle-Foot Orthosis Using Simulated Physical Human-Robot Interaction via Deep Reinforcement Learning

Adaptive state restricted barrier Lyapunov-based control of a Stewart platform used as ankle-controlled mobilizer

In this research project, a closed-chain robotic active ankle orthosis with six degrees of freedom is designed, constructed, numerically valued, instrumented, and experimentally validated. The mechanical arrangement to implement the orthosis corresponds to a six-legged Stewart platform. An adaptive gain control strategy with state constraints based on a state-dependent gains control (that behaves as a diverging function as the states approach the state restrictions) operates the device's motion. The convergence to an invariant positive set centered at the origin of the tracking error space is validated using the stability analysis based on the second method of Lyapunov, with the implementation of a state barrier Lyapunov-like function. The ultimate boundedness of the tracking error is proven with an endorsed gains adjustment method leading to a reachable minimum size of the ultimate bound. Hence, the impact of the state constraints and the formal reason for applying the controller on the suggested orthosis are all established. The orthosis is also controlled using a conventional state feedback strategy to assess the tracking error for an external disturbance and contrast its performance with the proposed control approach. The technology is tested on a few carefully chosen volunteers, successfully limiting the range of motion within a pre-defined region based on the scope of movement reported by patients with ankle illnesses discovered in the literature. Based on a unique mechatronic device, the created system offers a fresh approach to treating this class of impairments.

Shared Autonomy Locomotion Synthesis With a Virtual Powered Prosthetic Ankle

Virtual environments provide a safe and accessible way to test innovative technologies for controlling wearable robotic devices. However, to simulate devices that support walking, such as powered prosthetic legs, it is not enough to model the hardware without its user. Predictive locomotion synthesizers can generate the movements of a virtual user, with whom the simulated device can be trained or evaluated. We implemented a Deep Reinforcement Learning based motion controller in the MuJoCo physics engine, where autonomy over the humanoid model was shared between the simulated user and the control policy of an active prosthesis. Despite not optimising the controller to match experimental dynamics, realistic torque profiles and ground reaction force curves were produced by the agent. A data-driven and continuous representation of user intent was used to simulate a Human Machine Interface that controlled a transtibial prosthesis in a non-steady state walking setting. The continuous intent representation was shown to mitigate the need for compensatory gait patterns from their virtual users and halve the rate of tripping. Co-adaptation was identified as a potential challenge for training human-in-the-loop prosthesis control policies. The proposed framework outlines a way to explore the complex design space of robot-assisted gait, promoting the transfer of the next generation of intent driven controllers from the lab to real-life scenarios.